Framework
Three levers to raise screener predictive value
Ordered from strongest to weakest
Three ways to raise the share of people passing a research screener who really are your target, ordered by strength. Screening a pool where the target is common does far more than any rewrite. A second test only helps when it is independent, such as a behavioural check or an open walk-through of the last time someone did the task. Making questions harder to fake comes last. In Koji, an AI-moderated opening that asks for a real recent example can act as that second test.
Where this appears
What's in this framework
- 1. Raise the incidence of the pool: The strongest lever. Screening your weekly-active list instead of a general panel can take incidence from 3% to 60%.
- 2. Add a second, independent test: A behavioural check like telemetry, an artefact or an account lookup, not another self-report question in the same format.
- 3. Improve specificity, last: Questions that are harder to fake help, but move the number least and tightening has a cost of its own.
Use this image
Free to use in your docs, decks and posts under CC BY 4.0, as long as you credit Koji with a link to this page. To embed it, paste this into your page:
html
<a href="https://www.koji.so/visuals/three-levers-to-raise-screener-predictive-value"><img src="https://www.koji.so/visuals/three-levers-to-raise-screener-predictive-value/three-levers-to-raise-screener-predictive-value.png" alt="Framework of three levers to raise a screener's predictive value, strongest first: raise pool incidence, add an independent test, then improve specificity" width="1200" height="649" /></a>
<p>Source: <a href="https://www.koji.so/visuals/three-levers-to-raise-screener-predictive-value">Three levers to raise screener predictive value</a>, Koji</p>
